> For the complete documentation index, see [llms.txt](https://docs.aloop.icustomer.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.aloop.icustomer.ai/concepts/system-architecture.md). # System architecture How iCustomer Platform is put together: iHarness running the loop on your Growth Brain, the five layers underneath, and the four rules that keep your systems yours. iHarness runs on your Growth Brain. Everything else in the picture exists to feed that loop or to keep it honest. ![iCustomer Growth Brain: how it connects to your stack](https://4058268564-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FCtiCUwDtqLgVwntPjeix%2Fuploads%2Fgit-blob-3ef4585b69f2c4eb0bc98a2eab76bbf57a22efb4%2Fgrowth-brain-architecture.svg?alt=media) ## How to read it It is a loop, top to bottom and back. A goal comes in at any surface. iHarness runs the loop on the context the Growth Brain holds. Audience decides who. Decisions decide what. Every decision passes through the guardrails strip before iConnect carries it out in your systems. Measurement proves what worked, and outcomes flow back into the Growth Brain. That return path is the self-learning part: it is how the system works, not a feature label. ## The five layers | Layer | What it is | In one line | | ------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------- | | **1. Surfaces** | Where your team already works: the web app, Slack, the Snowflake native app, and agents connecting in. An access layer, not a product. | Works where your team works. | | **2. iHarness** | The agent running the loop. It holds three things: your goals and targets (without a goal, nothing runs), Pulse (the human channel: alerts, approvals, autopilot settings), and Orchestration (runs plays through agents, and plays live here). | The agent running your growth loop. It executes, within guardrails. | | **3. Growth Brain** | The self-learning context: customer, company, and audience, built on OneSource identity, accruing from every connected system and every decision outcome. iHarness's memory and the Growth Brain are the same thing. It holds three capabilities, covered next. | Everything you connect accrues into one self-learning context. | | **4. Guardrails & governance** | The strip between Decisions and iConnect: observability, evals, sandbox, decision traces. Every decision passes through it before touching your systems, and every outcome passes back through it. | Sandbox first, evals always, every decision traced. | | **5. iConnect** | Integrations. Activation out, governed connection in. Both directions, one layer. | Connects your stack: governed, both ways. | ## Inside the Growth Brain The third layer holds the three capabilities the loop is made of. They are the three boxes in the middle of the diagram. | Capability | What it does | In one line | | ---------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------- | | **Audience + Signals** | Who and when. FIRE scoring (Fit, Intent, Recency, Engagement), cohort building, and the first-, second-, and third-party signal feeds that drive them. | Decides who to act on and when, scored by FIRE. | | **Decisions** | The unit of work. Produced by plays, traced, and approved by you or by a policy you set. A specific, auditable action, reversible where the channel allows. | Every action is a traced, approved decision. | | **Measurement** | Did it work, causally. Two feeds, what each channel says happened and what iReveal proves happened, reconciled. | Causal proof, not last-touch attribution. | ## Below the platform: your stack Below the five layers sits your stack: your first-party systems (CRM, data cloud, drive, code, ESP, campaign tools), the media platforms where activation lands (LinkedIn, Meta, Trade Desk, Google), and 50+ third-party data vendors, resolved to OneSource IDs. Your data stays yours. Licensed data arrives resolved, never raw. The Growth Brain is not a CDP, a data lake, or a warehouse. It sits on your data infrastructure and does not replace it: the decision layer above the data cloud. {% hint style="info" %} **Common use cases.** Discover accounts and contacts always-on per ICP, monitor the intent of your ICP, prioritize on FIRE, route leads and govern contacts, and de-anonymize website visitors to prove outcomes. The plays that carry them are in the [Play library](/reference/play-library.md), and the visitor side is [iReveal](/ireveal/overview.md). {% endhint %} ## Where it runs * **Enterprise** runs in your cloud or warehouse. * **SMB** runs in our managed cloud. Same architecture, different home. "Zero egress" is the Enterprise shorthand, and it carries its scope: no warehouse data leaves except governed, customer-approved activations. Activation is egress by design. That is what you are running it for. ## Four rules the architecture enforces * **Only iConnect touches your systems.** Nothing writes around it. * **Every write to your systems is a decision**: traced, approved by you or by a policy you set, and reversible where the channel allows. * **Every decision passes the guardrails strip before it runs.** No exceptions for small actions. * **Third-party data never crosses raw.** Licensed data from the vendor layer enters only by resolving to OneSource IDs, and is never written raw into your CRM or media accounts. The missing arrow between the vendor layer and your systems in the diagram is deliberate: it is the compliance story. ## One diagram, three grains The loop runs at population grain over roughly 90 days: a commitment, from goal to outcome. [Data flow](/concepts/data-flow.md) runs at individual grain in seconds: one event, from capture to attribution. The five layers above are the machinery both run on. They meet at Audience, where one event becomes one identity, and identities become the population a KPI is a number about. ## Where to go next * [The loop in one page](/getting-started/the-loop.md): the path through these layers. * [Data flow and trace spine](/concepts/data-flow.md): the individual-grain path. * [Decisions and traces](/concepts/decisions-and-traces.md): what Decisions produce. * [Context and memory](/concepts/context-and-memory.md): how the Growth Brain ranks what it knows. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.aloop.icustomer.ai/concepts/system-architecture.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.